Retrieval practice: why low-stakes quizzes beat re-reading

Many pupils revise by re-reading their notes. It feels productive, but the research points to something that works better: trying to recall the material with the notes closed. This article sets out what the main studies found, with their numbers and their limits. By the end you will be able to plan a five-minute, low-stakes quiz that pupils mostly get right, that revisits older topics and that ends with the answers.

What the laboratory studies found

In a 2006 study, Roediger and Karpicke asked undergraduates at Washington University in St Louis to read short passages on science topics. The students then either read each passage again or wrote down everything they could remember, with no feedback.

In the first experiment, 120 students took a final recall test after 5 minutes, 2 days or 1 week. After 5 minutes, re-reading won: students recalled 81% of the ideas in a re-read passage, against 75% in a tested one. After 2 days, testing led by 68% to 54%, and after 1 week by 56% to 42%.

A second experiment, with 180 students, compared reading a passage four times with reading it once and recalling it three times. After a week, the repeated readers recalled 40% and the repeated recallers 61%. Yet the repeated readers had predicted that they would remember the passage better than the other groups.

A 2013 review by Dunlosky and colleagues rated ten study techniques. Only two were rated high utility: practice testing and distributed practice, because they 'benefit learners of different ages and abilities'. Re-reading and highlighting were rated low utility, and the authors recommend 'practice testing instead of rereading'.

Rowland's 2014 meta-analysis pooled experiments that compared testing with restudying. Tests that asked students to recall an answer produced larger benefits than recognition tests. Multiple-choice questions are usually classed as recognition, so mix them with questions that make pupils produce the answer.

What happened in real classrooms

Laboratory results do not always survive a real timetable. Roediger, Agarwal and colleagues (2011) worked with a US sixth-grade social studies teacher, so the pupils were roughly Year 7 age. Pupils took three short multiple-choice quizzes on half of the facts in each chapter, with the right answer shown after each question.

On the chapter tests, which counted towards grades, pupils scored 94% on quizzed facts and 81% on the rest. On an unannounced end-of-semester exam, the figures were 79% and 67%. These are the 36 pupils present at every session; adding those who missed sessions did not change the pattern.

A second experiment added re-reading. Some facts were quizzed, some were shown as statements to read and some were not revisited. Chapter test scores were 91%, 83% and 81%, and the re-read facts did not score significantly higher than facts that were not revisited. One caution: the multiple-choice chapter tests reused the quiz questions. The benefit also appeared when pupils wrote their own answers, in a free-recall test and in a third experiment with short-answer questions.

Two reviews pull the classroom evidence together. Agarwal, Nunes and Blunt (2021) coded 50 experiments in schools and classrooms, with 5,374 students in total. Of the 49 effect sizes, 57% showed medium or large benefits from retrieval practice. Only 6% of the experiments took place outside Western, educated, industrialised, rich and democratic countries.

Yang and colleagues (2021) combined 222 classroom studies covering 48,478 students. They conclude that quizzing 'raises student academic achievement to a medium extent', with an average effect of g = 0.499. Put simply, quizzed groups scored about half a standard deviation higher on average. The size of the benefit depended on factors such as corrective feedback, the number of quizzes and what the comparison group did instead.

What low-stakes means in practice

In these studies, low-stakes had a precise meaning. Roediger and colleagues used quizzes 'that count little or nothing toward a student's grade'. In the programmes Agarwal and colleagues surveyed in 2014, quizzes typically did not count toward grades, happened at least once a week and almost always gave immediate feedback.

That survey also asked about nerves. Of 1,408 pupils aged 11 to 18 in one US school district, 72% said the quizzes made them less nervous for unit tests and exams. Only 6% said more nervous. This is self-report from pupils who had taken part, so read it as encouraging rather than as proof.

Rosenshine's Principles of Instruction (2012) point the same way. His first principle is to begin a lesson with a short review of previous learning. He reports that the most effective teachers in the studies he drew on began with 'a five- to eight-minute review of previously covered material'. His tenth principle adds weekly and monthly review.

Why spacing and feedback matter

Spacing, which Dunlosky's review calls distributed practice, means returning to a topic after a gap instead of practising it all at once. That review rated it high utility alongside practice testing. The EEF's 2021 review of studies run in schools is more cautious. It found 18 studies of spacing across days and lessons, which suggest 'a small positive impact' on learning.

Results varied widely, and a significant number of those studies found no effect or a negative one. In only six of the 18 were pupils taught by teachers rather than by researchers or computer packages. The two reviews differ partly in scope: the EEF included only studies that took place in schools.

Feedback matters especially with multiple choice, because a wrong option that a pupil picks can stick. In a laboratory study with 72 undergraduates, Butler and Roediger (2008) gave a multiple-choice test with immediate feedback, feedback at the end or none. A week later, both kinds of feedback raised correct answers and cut the number of wrong options students repeated. Their advice: 'Educators should provide feedback when using multiple-choice tests.'

What the evidence does not show

The evidence is strong, but it has limits. Keep them in mind before you promise colleagues results.

• It does not replace teaching. The EEF advises that 'pupils should be able to retrieve at least some of the content they are tested on'. When pupils cannot retrieve the answer, the quiz 'is unlikely to increase memory strength'.
• Fact quizzes do not build reasoning by themselves. In Agarwal's 2019 study with middle school and college students, 'higher order and mixed quizzes improved higher order test performance, but fact quizzes did not'. The EEF also calls the evidence on higher-order learning uncertain.
• Material you do not quiz gains less. Pan and Rickard (2018) analysed 122 experiments on transfer. Testing helped with new questions to some degree, but least of all 'to untested materials seen during initial study'.
• The right difficulty is not settled. Rosenshine gives an optimal success rate of 'about 80 percent' during classroom instruction. The EEF says the evidence on the optimal level of challenge is uncertain.
• Few studies were run by ordinary teachers. Of the 21 school studies the EEF found comparing retrieval practice with restudy, only one was delivered by regular class teachers.

The EEF's overall view is that the positive results, the sound theory and the low cost make retrieval practice 'a promising approach that teachers should consider'.

Try it on Monday

Rosenshine suggests one way to build review into the timetable: 'review the previous week's work every Monday and the previous month's work every fourth Monday'. A five-minute starter can do that job.

1. Write five questions before the lesson: two on last week's topic, two on last month's and one from last term. For Year 8 science: 'Write the word equation for photosynthesis.'
2. Ask for recall first. Pupils write their answers on mini whiteboards or in the back of their books, with notes closed.
3. Keep one multiple-choice question for an idea pupils often get wrong: 'How does a plant get its food? A: it absorbs it from the soil. B: it makes it by photosynthesis. C: it takes it in from rainwater.'
4. Pitch the questions so that most pupils get most of them right. If almost nobody can answer one, reteach that point instead of quizzing it again.
5. Give the answers at once. Read out each correct answer and say why the most popular wrong answer is wrong.
6. Record nothing in the markbook. Pupils mark their own work and keep the score.
7. Note the two questions most pupils missed and put them into next week's quiz. Once a week, add one question that asks pupils to use what they know, such as 'A pupil says 0.3 × 0.2 = 0.6. Explain the mistake.'